Just to chime in here... I work with Mark... for the last year and half or so, 
under 2.7 - the exact same graph shows a steady bandwidth usage, and constant 
fairly fast sending. The ONLY difference here is the upgrade to 2.85 (same 
newsletters, tokenization, same machine, os, cf, firewall, network etc.). So 
obviously there is a difference in 'how' the process works... the question 
really is, are there settings somewhere that can be 'tuned' for better 
performance, or does the new methodology rely on possibly higher than 256 
threads to obtain similar steady performance - Howie? 

Is anyone else out there sending 50,000 tokenized 20k emails in under 1 hour 
smoothly using 2.8 on a 256 thread system?

Thanks,  

Jeff

Jeff Loechner
President, MediaPost Communications
[EMAIL PROTECTED]
212-204-2001

----- Original Message -----
From: "Stephen Garrett" <[EMAIL PROTECTED]>
To: "inFusion Support List" <[email protected]>
Sent: Friday, March 23, 2007 8:08:58 PM (GMT-0500) US/Eastern
Subject: Re: [iMS] Upgraded to 2.8, using half the bandwidth


Mark,

What a beautiful graph. Clearly processor used as disk i/o's occur, 
followed by network bandwidth increasing, then similar lags on the 
downhill side.

1st thing, is the "floating" seconds label should be for "MX Resolve 
Timeout" which defaults to 30 seconds. Looks like an oopsie there in 
your version of the configurator...

So what can cause the lags?

1) I assume you are using Cold Fusion for processing of templates? If 
so, in PERFMON there is the "Coldfusion Server" with some counters. I 
think it would be helpful if you could graph those as well. Perhaps 
CF is getting in the way in some fashion.

2) Your Simultaenous App Server connection is set to 5. This is the 
maximum number of simultaneous requests that the POST Server may make 
to the application server. Any POST requests over this number will be 
queued. Perhaps the lag is here?

3) Curious about your Allocation factor. The docs say it should be 
1/2 of your thread count. Did the 50 come from Howie?

Just trying to help...

At 04:04 PM 3/23/2007, Mark Kecko wrote:
>With the suggested settings its clear that there's some sort of 
>queuing or buffer problem, as suggested on the board. I disabled the 
>SMTP server and Scheduler and just kept POST running, did another 
>send, and took another Perfmon screen shot of my sawtooth bandwidth monster!
>
>Something, POST I'm assuming, kicks up the processor, it process 
>some mail, sends it, then takes a break, over and over.
>
>Mark
>
>Mark Kecko wrote:
>>I set my settings as described below, except for keeping
>>Sending IP Address: Auto
>>When I changed it to our internal IP our server wasn't connecting 
>>for SMTP conversations.  I switched it back to "Auto" and left all 
>>of the other settings as suggested. I got the same sawtooth 
>>behavior in my Performance Monitor as far as bandwidth is 
>>concerned, and slower sending as a result.
>>Thanks for your suggestions though!  I'd appreciate anyone else's 
>>input as well.
>>Mark
>>Gordon Burns wrote:
>>>At 19:30 23/03/2007 Ben Mueller said....
>>>
>>>
>>>>* You should not have "auto" as the IP address, but rather should 
>>>>select the specific IP address from which you want to send.  I 
>>>>was told by Howie that this can affect performance (though surely 
>>>>not at the level you're seeing)
>>>
>>>
>>>Just to chip in I agree with this and mine is set to the IP.
>>>
>>>
>>>
>>>>* My MX cache is set to 5000, not 50,000
>>>
>>>Agreed, 50,000 seems a very large number
>>>
>>>>* I, too, have a blank space with the "seconds" just kinda 
>>>>hanging out there.  I'm running 2.8.4 of the configuration tool.
>>>
>>>
>>>Yes me too.
>>>
>>>
>>>>* I have my "Maximum size of Memory Queue" set to 0, which I 
>>>>think means unlimited.
>>>
>>>Yes me too..
>>>
>>>
>>>
>>>>* I have my "minimum bandwidth per thread" set to 100, and yours 
>>>>is at 1000.
>>>
>>>
>>>me too
>>>
>>>
>>>>* I have my "simultaneous app server connections" set to 3.  I 
>>>>have a dim memory of this being important, but I don't remember 
>>>>exactly why. If I had to guess, I'd say that the number should be 
>>>>less than the number of CF threads you have available (assuming 
>>>>you're using CF).
>>>
>>>
>>>Me too
>>>
>>>
>>>>* I do *not* have "check all MX Servers" checked.
>>>
>>>Likewise
>>>
>>>I would be very interested to see if changing to the settings that 
>>>Ben and I share have any impact.
>>>
>>>Gordon
>>>PS Like Ben we do not see this performance issue that you are encountering.
==^=======================================================
This list server is Powered by iMS  "The Swiss Army Knife of Mail Servers"
--------------------------------------------------------------------------------------
This list is provided as a free service.  Although we will try to address issues
in a timely manner, support via this list is not guaranteed.  If you require 
expedited
support then a support contract is required.  Support may be purchased from
http://www.coolfusion.com/commerce.  Details regarding support options may be 
reviewed
at: http://www.coolfusion.com/SupportOptions.cfm
--------------------------------------------------------------------------------------
To leave this list please complete the form at 
http://www.coolfusion.com/Support/
Need an iMS Developer license?  Sign up for a free license here:
http://www.coolfusion.com/Developers/
List archives: http://www.coolfusion.com/cfbb/
Note: You are subscribed as archive_jab_org / [email protected]
==^=======================================================


Reply via email to